Ask any PI what worries them most about their lab's methods, and many will say the same thing. It isn't the protocol. It's the person who knows how to make it work.
When a senior graduate student defends or a postdoc moves on, years of hard-won technique can leave with them. The protocol stays in the shared drive. The judgment calls, the "watch for this at step 6," and the reasons behind the workarounds usually don't.
Nobody likes writing protocols
Everyone knows these gaps exist. People forget steps when they write things up. Even a notebook entry that a supervisor signs off on is rarely complete. That isn't carelessness. Experts stop noticing the things they do automatically.
Tech transfer has the same problem
In conversations with lab automation leaders, I've heard a version of this again and again. A team builds a complex assay by hand. Another team, or an instrument, has to take it over. The handwritten protocol arrives, and the gaps are enormous. Someone spends weeks rediscovering what the original scientist knew.
This will only become more common. As labs move methods onto instruments, and instruments gain shared languages like Anthropic's Model Hardware Standard, the bottleneck shifts to the human side. We need agreed terms for what a person did at the bench so it can be carried over to another person or to a machine.
Capture the expert while they work
The easiest time to record expert knowledge is while the expert is doing the work. Not in a write-up afterward, but in the moment.
One study I'd like to run looks like this:
Proposed study
- An experienced scientist runs a protocol normally, narrating out loud a little more than usual: what they're checking, what they adjust, and why.
- That run becomes a structured execution record.
- An incoming student uses the record to learn the same protocol.
- We measure time to mastery by how many known, predefined mistakes the new student avoids, compared with learning the usual way.
If the record transfers expertise, we should see it in how quickly new people get good.
Why a standard matters here
A record that only one app can read is just another silo. If execution records follow an open standard, the same record can train a new student, inform a tech transfer, and feed the translation into instrument methods. The knowledge stays with the lab instead of leaving with the person.